Dolly-v2 Models

Welcome to Dolly-v2 models! Created by Databricks, these AI models have been trained to follow instructions effectively. They were fine-tuned using around 15,000 instruction/response data elements provided by Databricks employees.

Dolly-v2 models showcase high-quality results in carrying out tasks based on specified instructions. While they may not represent the latest development in this field, the results they provide tend to surpass that of their foundational model.

This family of models includes varying complexities, such as dolly-v2-12b and dolly-v2-7b, each boasting billions of parameters.

These models have been trained using various data sources, including texts sourced from the public internet and a corpus specifically generated for their training. However, it’s important to note that the training data might contain limitations or biases inherent to the source’s characteristics.

Databricks continues to strive for ongoing research and development to create AI technologies that are helpful, honest, and harmless, truly aiming to maximize potential.

For more technical information or tips on running these models, consider visiting the Dolly GitHub repo .